Why Replacing Developers With AI is Going Horribly Wrong — Transcript

Full transcript

1. 0:00Tesla, Microsoft, Amazon, Google, and
2. 0:02almost every tech company you can
3. 0:04imagine have tried to replace humans
4. 0:05with AI.
5. 0:06>> This week Microsoft announcing it is
6. 0:08laying off nearly 4% of its workforce.
7. 0:10The CEO of Amazon telling the company's
8. 0:12employees that some of their jobs will
9. 0:14be replaced by AI.
10. 0:15>> In recent years, companies have laid off
11. 0:17huge numbers of workers.
12. 0:19>> The new year is not starting off on a
13. 0:21happy note for workers at some of the
14. 0:22country's largest tech companies.
15. 0:24>> This was because companies were
16. 0:25overconfident about their AI and
17. 0:27automation developments.
18. 0:29>> Amazon is warning its employees that the
19. 0:31company will have a smaller workforce in
20. 0:33the future because of generative AI and
21. 0:35agents.
22. 0:36>> However, it seems that companies are
23. 0:38quietly backpedaling as according to
24. 0:40OrgView of senior business leaders and
25. 0:42executives, 55% of businesses that
26. 0:45replaced employees with AI regret it.
27. 0:47>> Tesla, Klarna, IBM, Duolingo, and a
28. 0:50countless more. They were overconfident
29. 0:53in replacing humans, but now most are
30. 0:56quietly backpedaling.
31. 0:57>> So, why tech is regretting replacing
32. 0:59humans with AI?
33. 1:02Companies' attempts to automate
34. 1:04processes with minimal human
35. 1:05intervention are not new. Tesla's case
36. 1:08in 2017 is an early example of the
37. 1:11limits of extreme automation. The
38. 1:13company sought to build the so-called
39. 1:14machine that builds the machine with a
40. 1:16production workshop almost entirely
41. 1:18automated for the Tesla Model 3.
42. 1:20>> Elon Musk, Tesla's co-founder and CEO,
43. 1:23has referred to the Gigafactory as the
44. 1:25machine that builds the machine, and
45. 1:28it's all part of his master plan to make
46. 1:30electric cars more affordable.
47. 1:32>> goal was to achieve the production of
48. 1:335,000 vehicles per week, but results
49. 1:35were significantly lower due to
50. 1:37machinery failures, production
51. 1:38bottlenecks, and systems that did not
52. 1:40operate as intended. Robots specialized
53. 1:43in tasks such as placing front seats
54. 1:45broke down up to five times a day, while
55. 1:47industry standards show operational
56. 1:49intervals exceeding a month without
57. 1:51interruptions. The combination of new
58. 1:52technologies and lack of preventive
59. 1:54maintenance turned the production line
60. 1:56into a fragile system. Tesla had to
61. 1:58reintegrate human staff through a
62. 2:00temporary line called the Sprung
63. 2:02Project, where workers replaced the
64. 2:04robots.
65. 2:05>> In April, Musk tweeted that, "Quote,
66. 2:07humans are underrated." He was referring
67. 2:09to Tesla's experience making its new
68. 2:11Model 3s with one of the most
69. 2:12robotics-dependent assembly lines on the
70. 2:14planet. That grand experiment failed to
71. 2:16deliver nearly the number of cars Tesla
72. 2:19promised.
73. 2:19>> Production eventually accelerated,
74. 2:21allowing the company to avoid
75. 2:22bankruptcy. But, this example shows that
76. 2:25the efficiency promised by automation
77. 2:27can generate organizational fragility
78. 2:29when the value of human labor is
79. 2:31underestimated. With the adoption of
80. 2:32advanced language models, companies
81. 2:34began applying this logic to the
82. 2:36replacement of intellectual and service
83. 2:38tasks, including customer service,
84. 2:40marketing, data analysis, and content
85. 2:42development. Klarna, for example,
86. 2:44implemented chatbots that replaced
87. 2:46hundreds of agents, reducing its
88. 2:48workforce from 5,000 to 2,000 employees.
89. 2:51Initially, the company reported that
90. 2:53chatbots managed 2/3 of interactions,
91. 2:55but later a decrease in service quality
92. 2:57was identified.
93. 2:58>> In 2023, the CEO of Klarna said AI would
94. 3:01replace half his workforce in just a few
95. 3:03years. And Klarna eventually reduced its
96. 3:06workforce by 40%, but in 2025, Sebastian
97. 3:09flipped that narrative on its head. He
98. 3:11publicly stated that quality human
99. 3:13support is the way of the future for us.
100. 3:16>> In fact, leaked internal data indicate
101. 3:18that problem resolution times increased
102. 3:20by 27% while unsatisfactory interactions
103. 3:23grew by 35% in the first 3 months after
104. 3:26implementation. Failures included
105. 3:28incorrect approval of leave requests and
106. 3:30inadequate to internal conflicts. This
107. 3:33showed that AI is more effective in
108. 3:35structured tasks than in processes
109. 3:37requiring contextual judgment.
110. 3:39Similarly, Taco Bell experimented with
111. 3:41an automated voice system in 500
112. 3:43locations. However, the company had to
113. 3:46limit its implementation due to errors
114. 3:48in orders and billing, which affected
115. 3:50customer experience and operational
116. 3:52efficiency.
117. 3:52>> I don't know if you either of you two
118. 3:54have interacted with an AI chatbot
119. 3:56customer service thing, but it is a
120. 3:58nightmare. I was stuck in like a chatbot
121. 4:00death loop of like trying to get it to
122. 4:02do what I wanted. Eventually, I had to,
123. 4:04you know what, pick up a phone and call
124. 4:06a human being.
125. 4:07>> Duolingo implemented its system called
126. 4:09AI first to replace part of contractors'
127. 4:11work, aiming to improve efficiency and
128. 4:13reduce costs. However, the company later
129. 4:16reported a noticeable decline in lesson
130. 4:18quality with errors affecting up to 42%
131. 4:21of content in some courses and causing
132. 4:23an 18% drop in user retention during the
133. 4:26first quarter after the changes. Another
134. 4:28example is the Australian company
135. 4:30Telstra, which replaced 2,800 employees
136. 4:33with AI, but saw customer response times
137. 4:36increase by up to 25%. Shopify
138. 4:39conditioned the hiring of new employees
139. 4:41on proving that AI could not perform
140. 4:43certain tasks, generating project delays
141. 4:45and an internal environment of
142. 4:46uncertainty. MIT analyses indicate that
143. 4:49only 5% of AI integrations generate
144. 4:52immediate revenue increases. The
145. 4:54remaining 95% failed to achieve
146. 4:56significant results, mainly due to
147. 4:59insufficient planning, lack of adequate
148. 5:01data, and inadequate employee training
149. 5:03in the use of these tools. Tech giants
150. 5:05like Microsoft and Google are
151. 5:06outsourcing more and more coding to AI
152. 5:09in a productivity push, but some new
153. 5:11research shows the tools might not be as
154. 5:13helpful as some expect.
155. 5:14>> Organizational impact is also reflected
156. 5:16in employee turnover. Companies that
157. 5:19implemented AI without human supervision
158. 5:21experienced an average increase of 22%
159. 5:24involuntary turnover during the first 6
160. 5:27months, raising recruitment and training
161. 5:29costs by 18%. Additionally, customer
162. 5:32experience was affected with decreased
163. 5:34satisfaction and loyalty metrics. The
164. 5:36fragility of automation is evident when
165. 5:38a small AI breakdown can halt entire
166. 5:41operations, while human intervention can
167. 5:43resolve incidents immediately. Despite
168. 5:45these challenges, AI can be
169. 5:46complementary if implemented
170. 5:48strategically. 80% of leaders plan to
171. 5:50train their employees in AI tools, and
172. 5:5341% have increased their learning and
173. 5:55development budgets.
174. 5:56>> And we thought that it was really
175. 5:58important that we help upskill people.
176. 6:01So, we are investing a lot into this.
177. 6:04>> Startups and companies that apply AI
178. 6:06gradually and purposefully report
179. 6:08productivity increases of up to 35% and
180. 6:11operational cost reductions of 27%. This
181. 6:14shows that the combination of AI with
182. 6:16human supervision produces better
183. 6:17results. In logistics, for example,
184. 6:20route optimization with AI accompanied
185. 6:22by human supervision has reduced
186. 6:24delivery delays by 18% without
187. 6:27compromising customer experience. But,
188. 6:29the implementation of AI in companies
189. 6:31has shown mixed results in terms of
190. 6:33employee turnover. On one hand,
191. 6:35automating repetitive tasks can free
192. 6:37employees for more strategic roles,
193. 6:39reducing burnout and improving
194. 6:40retention. On the other hand, the
195. 6:42perception that AI can replace jobs can
196. 6:44generate job insecurity and increase
197. 6:46turnover.
198. 6:47>> A new article from Business Insider
199. 6:48describes so-called office paranoia, and
200. 6:51it outlines how factors like artificial
201. 6:54intelligence and changes to the labor
202. 6:55market are bringing a sense of dread to
203. 6:58workplaces all over the place.
204. 7:00>> Furthermore, leaders recognize that AI
205. 7:02implementation can generate financial
206. 7:04and operational benefits. However, its
207. 7:06effectiveness depends on comprehensive
208. 7:08planning that considers training,
209. 7:10supervision, quality protocols, and
210. 7:12adaptation to existing processes. AI is
211. 7:15also proving not to be as effective on
212. 7:17its own. A MIT study revealed that only
213. 7:207% of AI initiatives in companies
214. 7:23generate a significant return, while the
215. 7:25remaining 93% produce no measurable
216. 7:28results. This finding underscores the
217. 7:30importance of strategic and well-planned
218. 7:32AI implementation, considering the
219. 7:34specific needs of the company and proper
220. 7:36integration with existing processes.
221. 7:38Evidence suggests that AI works best
222. 7:40when it complements human skills rather
223. 7:42than attempting to fully replace them.
224. 7:44This approach allows employees to focus
225. 7:46on higher-value strategic tasks, while
226. 7:49automated systems handle repetitive or
227. 7:51large-scale analytical functions.
228. 7:53>> A new poll shows more workers in the US
229. 7:55are using artificial intelligence to
230. 7:57help free up hours at work. The survey
231. 7:59from ResumeBuilder shows 40% of people
232. 8:03using ChatGPT say they save 1 to 5 hours
233. 8:06per week.
234. 8:07>> Data also show that internal perception
235. 8:09of AI affects adoption and results. When
236. 8:12employees perceive that their work is
237. 8:13devalued, the risk of stress,
238. 8:15demotivation, and talent loss increases.
239. 8:18Therefore, responsible AI integration
240. 8:20requires clear communication about
241. 8:22objectives, benefits, and limitations of
242. 8:24the technology, as well as continuous
243. 8:26training programs. Evidence from
244. 8:28companies such as Tesla, Klarna, and IBM
245. 8:31demonstrates that organizational
246. 8:32resilience depends on the balance
247. 8:34between technology and human capacity.
248. 8:37[Music]
249. 8:39At Economy Media, your opinion matters
250. 8:41to us. Subscribe and let us know what
251. 8:43you think in the comments below.